Skip to content

KwaiAgents

Open-source agent models, data and benchmarks

Visit official site
Category AI Agents

What KwaiAgents is

The project is described as a series of open-source agent works released by KwaiKEG at Kuaishou. It bundles a light agent system, language models, instruction tuning data and a human evaluation benchmark.

What you can do with it

  • Use agent language models with planning and tool use
  • Work with large-scale agent instruction data
  • Evaluate agents on the released benchmark

Application scenarios

  • Chat about a film with an agent
  • Book a train ticket through a dialogue
  • Plan a driving route from an office to an airport

Pros & cons

✓ What we like

  • Models, instruction data and a benchmark are released together
  • Planning, reflection and tool use are listed abilities

! What to watch out for

  • The released system is described as a light version
  • Evaluation entries come from human assessment data

FAQ

What is included?

KAgentSys-Lite, KAgentLMs, KAgentInstruct and KAgentBench are named.

What does the benchmark test?

Planning, tool use, reflection, summarisation and description abilities are listed.

Last reviewed: 2026-11-15

More AI office assistant tools

View all →

How we review